automatic computing machinery 167 UNPUBLISHED MATHEMATICAL TABLES In this issue there is a reference to an unpublished table in RMT 1015 147[A, P].—Leo Storch, Admittance-Impedance Conversion Tables, Tech- nical Memorandum No. 274, Hughes Aircraft Co., Research and Devel- opment Laboratories, Culver City, California. 10 p. manuscript tabu- lated from punched cards. Copy deposited in UMT File. The table gives 4S values of (1 + s2)"1 and s(l + s2)-1 for 5 = 0(.001)1. It is intended to facilitate the calculation of the reciprocal of a complex number. The table is an extension of a table of Jahnke & Emde. [4th ed., appendix, p. 13. j 148 [F].—F. Gruenberger, Lists of primes. Six sheets tabulated from punched cards. Deposited in the UMT File. These lists of primes are for the ranges 10100009 to 10132211 and 50000017 to 50040013. The lists were computed on a CPC as a fill-in project, without attempting to program for speed. A graph showing the distribution of differences between consecutive primes in the ranges 1000003-1024523, 10100009-10132211and 50000017-50012839 is included. F. Gruenberger University of Wisconsin Madison, Wisconsin 149 [F].—A. Gloden, Table of solutions of the congruence a;128+1=0 (mod p) for p < 20000. Manuscript, 2 p., deposited in the UMT File. The table gives for each of the 16 primes p of the form 256¿ + 1 less than 20000, the 64 solutions of the congruence mentioned in the title which are less than \p. A. Gloden 11 rue Jean Jaurès Luxembourg AUTOMATIC COMPUTING MACHINERY Edited by the Staff of the Machine Development Laboratory of the National Bureau of Standards. Correspondence regarding the Section should be directed to Dr. E. W. Cannon, 415 South Building, National Bureau of Standards, Washington 25, D. C. Technical Developments AN AUTOMATIC COMPUTER IN AUSTRALIA An automatic computer, the "C.S.I.R.O. Mk. I Digital Computer," designed and constructed by the Radiophysics Division of the Common- wealth Scientific and Industrial Research Organization in Sydney, Australia, is now in service. It is of the all-electronic, serial-binary type with a main store consisting of a group of ultrasonic delay lines with a total capacity of 1,024 words of 20 binary digits each. An auxiliary store in the form of an unsynchronized magnetic drum is incorporated with a capacity of 1,024 similar words, later to be extended to 4,096 words. No attempt has been made to obtain very high speeds of operation, the 168 AUTOMATIC COMPUTING MACHINERY objective being to construct a computer which is simple and sufficiently flexible from the engineers' and mathematicians' point of view. Thus, equipment can be removed or added without dislocating the mode of operation. The device is intended at this stage primarily for research into numerical methods and programming techniques. Standard electronic components are used throughout the 1,500 or so tubes and circuits. The clock pulse rate is 333 kcs. A minor cycle of one word length covers a period of 60 ^sec, and the major cycle of 16 words, equal to the delay of each ultrasonic storage channel, covers a period of 1 millisecond. On the average the time occupied by the selection of a com- mand and the corresponding operation is about 2 major cycles. Numbers and Commands. Numbers are registered in the straight binary scale, with each of 19 digits together with a sign digit in the most significant place, and with negative numbers being stored in complementary form. The convention with regard to the position of the binary point is only significant in the operation of the automatic multiplier, where it is placed immediately to the right of the sign digit of the total product. Although each number consists of only 20 binary digits, it is not difficult to programme for 40 digits or even greater accuracy. The scheme adopted for commands is of the "two address" type, in which each operation is considered as a transfer of the content of one register to another, the former being specified as a "source" and the latter as a "destination." An arithmetical function is specified as a quality of the par- ticular transfer demanded by a command. A command is divided into three groups of digits. Two adjacent groups of 5 digits define the source and destination, and a further group of 10 digits specifies a numerical component normally used to indicate which store position is called for when either of the other two addresses involve the store. If the store is not called, the third address may be used to store special information. These digit groupings are in the order in which digits are transmitted from a register: destination, 1-5; source, 6-10; numerical, 11-20. Commands are held in and are normally accepted from the store in serial order. Organization. The computer is of the serial type, and all transfers take place along a single "digit trunk." All registers are connected to this trunk via "function gates" which are under the control of the central sequencing unit and the command decoding devices. The digit trunk consists of two parts, an "output trunk" and an "input trunk," and transfers are made between these conductors during single minor cycle periods, determined by a time selector which operates upon the detection of equality between digits 11-14 of the current command and the current minor-cycle number. The selection of a single command from the store, and its performance, requires four transfers. The registers involved in controlling these actions are: (A i) The "sequence register": a 10-digit register which keeps a tally of the progress of the programme, and instructs whence the next command is to be withdrawn. Its contents are normally increased by unity following the selection of a command. (A ii) The "store control register," of 10-digit capacity, which is con- nected in the upper 6-digit positions, i.e. (15-20), to a decoding selector AUTOMATIC COMPUTING MACHINERY 169 used to specify which store channel is needed, and in its lower 4 digit positions (11-14) to the time selector. (A iii) The "interpreter register," of 20-digit capacity, which is con- nected to the "source and destination selectors" each possessing 32 outputs. Some of these registers possess special function gates and may be called to transfer or receive by a command, as follows: (B i) The sequence register can be called to "count in," i.e., to increase its content by a number equal to the number of unit digits entering; to "add in" on the digit positions 11-20; and to "substitute in" on the same positions, i.e., reset and add in. Also it can be called to transmit in the digit group 11-20. (B ii) The interpreter register can be called to transfer out its numerical content on digit positions 11-20. The four transfers involved in satisfying a single command are of invari- able form. They constitute what is called the computer routine and are as follows : (C i) The content of the sequence register is transferred to the store control register whose content it replaces. (C ii) Under the control of the content of the store control register, the store is allowed to transmit. The time selector finally allows the desired command to enter the interpreter register via the "input trunk." (C iii) The command may call for the store to transmit or receive, hence the numerical part of the command is transferred from the interpreter to the store control register, replacing any previous content. (C iv) Under the control of the time selector and the action of the source and destination selectors, a selected pair of function gates is actuated, and the desired transfer is performed. The function gates and their actions are listed in Table I. The Arithmetical Unit. The arithmetical unit consists of a group of 5 ultrasonic delay-line registers of which A, B, C are of 20-digit capacity, whilst H is of 10 digits only. The fifth register D can store sixteen 20-digit numbers. Registers A, C, and D can add, subtract, and replace, whilst H can read in and replace in either of the digit groups 1-10 or 11-20. Register B is a non-adding register. Register A is the main accumulator and possesses a number of special gates for performing the functions of shifting right and left by one place, as well as reading out the most significant digit and the lowest digit. It is also capable of certain logical functions. Registers C and D are capable of reading out the sign digit of their content. This applies to any single number in D. Registers A, C, and D can read out their contents, and also may do so with simultaneous reset. Special functions associated with multiplication have been introduced which also may be used for other purposes. For automatic multiplication, registers A, B, and C are coupled together. Register C holds the multipli- cand whilst B receives the multiplier. The registers A and B are connected into series circulation together with an extra digit period, for a period of 41 minor cycles during which the multiplier digits are investigated succes- sively and removed from circulation. The total product is built up by successive addition and shifting, until actually it occupies registers A and 170 AUTOMATIC COMPUTING MACHINERY Table I. Function gates Command addresses: n; S, D Sources (5) Destinations (D) Read Out: Read In: (1) High speed store position "n" (1) High speed store position "n" (2) low speed store position "n" (2) low speed store position "n" (3) current command positions 11-20 (3) shift to binary/decimal reading (4) read current card column and shift (4) to output "print" to next column Í5) position of next command (5) to output and "punch"
Details
-
File Typepdf
-
Upload Time-
-
Content LanguagesEnglish
-
Upload UserAnonymous/Not logged-in
-
File Pages24 Page
-
File Size-